Preparing for Power Over Ethernet Plus

The original 802.3af PoE standard offered a fairly straightforward way to supply loads with 13 W or so of usable power delivered at 48 V dc. But IEEE 802.3at PoE Plus, which ups usable power to something over 50 W, introduces some wrinkles that designers and even IT managers must understand.

One catch is that designers can still supply power in a limited fashion in some existing Ethernet installations via a midspan bridge. But in that case, designers can’t implement power negotiations between a powered device (PD) and power source equipment (PSE). This implies dedicated PoE Plus ports and relatively high duty-cycle power supplies in midspans.

Something else to watch out for are PDs that dynamically negotiate power requirements with the PSE via their Ethernet connection. This requires more code in the PD microcontroller and a greater understanding of dynamic power requirements on the part of the engineer writing that code.

A potential pitfall for end users is that PDs can meet the standard by operating in a fall-back mode if there’s not enough power for full functionality. (For example, a video phone could fall back to operating voice-only, without a video display.) Alternatively, a PD application could meet the standard simply by signaling “insufficient power.” IT managers who bought a lot of “compliant” video phones could find themselves embarrassed by a system that didn’t work as expected if a “compliant” switch didn’t possess a sufficiently robust power supply.

To get comfortable with PoE Plus, it helps to understand its genesis and subsequent evolution. In the beginning, Cisco had a proprietary approach for powering Voice over Internet Protocol (VoIP) business phones that involved powering some pairs in the router with 48 V.

The rest of the industry saw that this was good and wished for an open standard, which became IEEE 802.3af. To be conservative, the IEEE subcommittee limited power to 15 W at the PSE, which was enough for the non-video VoIP phones that then dominated the market. They also expanded Cisco’s idea by allowing the “spare pairs” in an Ethernet cable to be powered by a midspan, making it possible to retrofit PoE to legacy Ethernet plant.

When PoE hit the streets, many potential vendors saw its advantages and jumped on the bandwagon. VoIP phones would no longer need power plugs, making them more like old-fashioned public-branch-exchange (PBX) phones. Wireless hotspots could be located anywhere someone could pull a CAT5 cable.

Supermarket shelves would twinkle with up-to-date price tags that would always match the prices in the cash register. And, PoE musical instruments, mixers, and recording equipment would displace the MIDI bus and revolutionize the music business.

Obviously, some of these goals were more realistic than others. In the three years since basic PoE was released, three killer applications have taken hold: VoIP phones, Wi-Fi hotspots, and security cameras. Within those applications, though, there immediately appeared a need for power beyond 13 W.

For example, there’s an anticipated demand for video conferencing using VoIP phones, and backlighting a video screen takes power. Simple short-range Wi-Fi is happy with 13 W, but WiMAX takes more power. And while fixed security cameras don’t require much power, once motors are added for panning, tilting, and zooming, power does become an issue.

But the manufacturers of Ethernet switches, concerned about over-specifying power supplies, pointed out that video phones and pan/zoom/tilt cameras don’t need full power all of the time. Most of the time, the phone is just sitting there. Even when there’s a call, video isn’t always necessary. Unless it’s a formal conference, most people would prefer to remain invisible to the other party. Similarly, those high-end security cameras only move when a guard touches a joystick. In other words, the requirement for higher power changes continuously.

The dynamic-power issue transformed the questions facing the IEEE 802.3at task force from simply “How much current can a bundle of CAT5 cables and their associated RJ45 connectors safely handle?” to “How can we create a protocol that allows PDs to dynamically negotiate for power with a PSE?”

Continued on Page 2.

Before we get into that, let’s take a look at basic PoE as described in the IEEE 802.3af, DTE Power via MDI, which was formally approved on June 12, 2003 (see the figure). PoE uses either the data pairs or the spare pairs in an Ethernet cable to carry 48 V dc from the PSE in an endpoint switch or midspan hub to the PD appliance at the other end of the cable. Data pairs are powered via center tap, while spare pairs are simply paralleled. The sense of the dc voltage doesn’t matter, thanks to a diode bridge ahead of the PD controller chip.

To take advantage of Power over Ethernet, PSEs must be able to detect the presence of a PD on any port. And, PD appliances must be able to assert their PoE compatibility and may assert their maximum power requirements. Under the 802.3af standard, PSEs may not apply power to the Ethernet cable unless there’s a PoE-enabled PD on the other end. PoE PDs are identified by the presence of a 25-kΩ resistor across their input.

The PSE measures resistance by applying two voltages (separated by 1 V and a 20-ms interval) and using the resulting currents to determine the resistance value. This part of the handshake is called the “discovery” phase. Next, there’s an optional “classification” phase. In the 15-W maximum world of basic PoE, classification allows the PSE to decide whether it has enough capacity to supply the PD. If it doesn’t, it can refuse to power up the Ethernet pair.

During the classification phase, the PSE briefly asserts a 15.5- to 20-V pulse on the pair, and the PD can opt to signal the PSE by placing a load on the line. Doing nothing, not putting a load online, automatically identifies the PD as Class 0, and the PSE would expect it to limit current to 400 mA. Class 1 PDs must selflimit to 120 mA, Class 2 to 210 mA, and Class 3 to 310 mA. A Class 4 was reserved for future use. Timing relationships were 500 ms (max) for detection, 10 to 75 ms for classification, and 400 ms for power turn-on.

Originally, PoE was intended for standard Ethernet cable, which has four twisted pairs. But only two of those pairs carry data. Under basic PoE, powering is an either/or situation—only one pair may be used at a time. This enables the seamless use of new endpoint routers with built-in PoE or power via midspan bridges in legacy systems. Midspans would only power the spare pairs, while endpoint equipment could power either pair. (In practice, all endpoints power the data pair.)

That arrangement eliminates the possibility of midspan PoE for legacy plant in which the spare pairs are left unconnected. However, the expectation was that most PoE would be endpointpowered in the long term. So, endpoints should be able to handle any kind of infrastructure, including legacy sites with unconnected spare pairs.

When using the spare pairs in basic PoE, pins 4 and 5 are paralleled for one side of the dc supply and pins 7 and 8 are paralleled for the other side. When using the data pairs, the PSE applies dc power to the center tap of each isolation transformer so that pins 3 and 6 supply one side of the dc and pins 1 and 2 supply the other. At the PD, datapair power is recovered via center taps on each transformer.

With that background, it’s possible to understand PoE Plus. Part of the IEEE 802.3at Task Force’s job was to decide whether the additional power would be delivered by simply increasing the maximum current rating or by paralleling the spare pairs with the data pairs. The more challenging part involved expanding the classification scheme for PDs to dynamically negotiate with the PSE for more or less power.

Resolving the first issue was relatively simple, once it was agreed that CAT5 cable and RJ45 connectors could handle more current and that was acceptable to use both sets of twisted pairs. One could have it both ways: more current and four active pairs. That decision impacted midspan makers, though. If there’s no continuity through the spare pairs, they can only supply half as much power as an endpoint switch.

That leads to a situation in which PoE Plus-compliant midspans can be configured with a mix of basic PoE, PoE Plus, and non-PoE ports—without the versatility inherent in a full PoE Plus endpoint. This isn’t a big disappointment to midspan makers, however. There was never going to be a way to deliver as much power through two pairs as was delivered through four. All they ever wanted was to deliver more than 13 W on dedicated ports at prices below those for new endpoint switches or routers.

So, what’s the real maximum power that can be supplied by PoE Plus? “The IEEE 802.3at current has been established at 360 mA per conductor, 720-mA delivered current by the TIA TR-42 working group. This current is good for up to a 45°C environment and must be de-rated to 0 mA at 60°C,” says Clay Stanford of Linear Technology.

“There are concerns about the 45°C and the de-rating. Because of this, it would be my opinion that the 720-mA current limit isn’t set in stone. It might be reduced to something like 500 mA so that a higher ambient temperature could be allowed,” he notes.

With respect to voltage, Stanford says, “The IEEE 802.3at committee has tentatively established the PSE output voltage as 50 to 57 V. The committee has established the total round-trip 100-meter cable resistance to be 12.5 x max.” With this voltage and resistance, the power is:

Ppse min = 0.72 x 3 50 V = 36 W
Ppd min = 0.72 x 3 41 V = 29.5 W
Cable loss max = I2R = 0.72 x 0.72 x 12.5 = 6.5 W max

The challenge for the 802.3at Task Force was to determine how to extend the classification system. It came up with a two-level classification system...

TAGS: Components
Hide comments


  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.